LessonThis is a lesson restated in new words. Any figure in it is an example from the source, not this desk's record.
A backtest is an unconditional game until you ask what everyone else is doing. In the lecture's poker picture, always continuing with pocket aces against five players who also continuation-bet is a positive-expectation simulation over 200 hands. Played against people who actually bet big hands, the same entry loses. Folding when the action gets large after the flop cuts the average loss enough to make the expectation positive again, even though folds add to the loss count.
Why it matters
An entry signal, in trading or in cards, is a rule for when you put money in. Pocket aces are the lecture's stand-in for something like a moving-average crossover. Expectation splits into the probability of winning, the probability of losing, the average win, and the average loss. Positive expectation means the winning side outweighs the losing side in that weighted sum. The 200-hand simulation, with static betting and no bluffs, produced an equity curve the lecture describes as positive expected value.
Take that signal "live," meaning into a game where opponents bet real hands, and the curve stops making money. The realized gain distribution does not match the backtest or the unconditional simulation. Some tables are the static game. Some are not. The lecture's instruction is to press where the edge is larger and to fold or hedge where it is smaller.
The idea in plain words
The live-style distribution is dominated by losses, which is what negative expectation looks like after you separate wins and losses and attach their probabilities. You can push the sum back up by raising the average win, raising the win probability, cutting the loss probability, or cutting the average loss. The lever pulled here is the average loss. The rule is to fold when someone bets big beyond the flop, instead of continuation-betting every street through the flop, turn, and river.
After that change, the same pocket-ace entry shows a more positive expectation in the lecture's live-market picture. The win and loss shapes can look similar. The probability of a loss, counting folds, actually rises. The average loser falls by enough to flip the sign of the expectation. Volume, volatility, and sentiment regimes are the trading names for the same idea: some entries are the table where you continue, and some are the table where you fold.
Risk, in this account, is deviation from the expected value. Sharpe and Sortino ratios are described as measurements of that deviation. They do not make two uses of the same signal into one distribution. A moving-average crossover at one date can be a winner and at another date a loser because the conditional situations differ, just as pocket aces differ between a table of automatic continuation bets and a table where someone bets up a real hand after the flop.
How to use it
Condition the signal on the state you can see: the broad climate, the volatility regime, volume, the spread. The lecture's claim is that when you do this well, each regime's gain distribution in the backtest stays recognizable in a walk-forward or live sample. Some regimes will show more positive expectation than others. Stability means the distribution you condition on is the distribution you later experience, so the equity curve can track a positive expectation. The lecture explicitly sets bet sizing and ergodicity aside. The point is the conditional expectation, not the stake.
What usually goes wrong
Shipping the unconditional simulation is the basic miss. Nobody bluffing and nobody holding a better hand is a different game from the one you will sit in. A second miss is to see the loss frequency rise after you add folds and conclude the rule failed. In the lecture's numbers, that frequency rose and the expectation still improved, because the average loss fell further.
Questions people ask
What was positive in the 200-hand simulation?
Always continuing with pocket aces against five opponents who bet the same automatic way. The lecture calls that curve positive expected value. It assumes no bluffs and no one betting a better hand.
How did the losing version get fixed?
By folding when someone bet big after the flop. Loss frequency, including folds, went up. The average loss fell enough that expectation turned positive.
What is the trading version of the fold?
Skip or hedge the entry when the regime, volume, volatility, or spread says you are in the situation with the worse conditional distribution, and keep the situation with the better one.